分治法实验夜间图像去雾 matlab代码
时间: 2023-08-27 19:30:10 浏览: 57
以下是一个简单的使用分治法实现夜间图像去雾的MATLAB代码:
```matlab
function output = dehaze(input)
% 分割图像为小块
patch_size = 15;
stride = 5;
[h, w, ~] = size(input);
output = zeros(h, w, 3);
count = zeros(h, w, 3);
for i = 1 : stride : h - patch_size
for j = 1 : stride : w - patch_size
patch = input(i : i + patch_size - 1, j : j + patch_size - 1, :);
% 计算暗通道先验
dark_channel = min(patch, [], 3);
A = prctile(dark_channel(:), 99.9);
% 估计大气光值
[x, y] = find(dark_channel >= A);
[~, idx] = max(patch(sub2ind(size(dark_channel), x, y, ones(size(x)))));
J = patch(x(idx), y(idx), :);
% 估计透射率
eps = 1e-3;
t = 1 - eps * min(patch ./ repmat(J, [patch_size, patch_size, 1]), [], 3);
% 修复颜色失真
t0 = 0.1;
patch_t = repmat(t, [1, 1, 3]);
patch_J = repmat(J, [patch_size, patch_size, 1]);
patch_A = repmat(reshape(A, [1, 1, 3]), [patch_size, patch_size, 1]);
patch_t = max(patch_t, t0);
patch_out = (patch - patch_A) ./ patch_t + patch_A;
% 合并图像块
patch_out(isnan(patch_out)) = 0;
patch_out(isinf(patch_out)) = 0;
output(i : i + patch_size - 1, j : j + patch_size - 1, :) = output(i : i + patch_size - 1, j : j + patch_size - 1, :) + patch_out;
count(i : i + patch_size - 1, j : j + patch_size - 1, :) = count(i : i + patch_size - 1, j : j + patch_size - 1, :) + 1;
end
end
% 取平均值
output = output ./ repmat(count, [1, 1, 3]);
output(isnan(output)) = 0;
output(isinf(output)) = 0;
output = im2uint8(output);
end
```
该代码将输入的夜间图像分割成大小为15x15的小块,并对每个小块进行去雾处理。具体地,该代码实现了以下步骤:
1. 计算每个小块的暗通道先验,即选取RGB三个通道中最小值。
2. 估计每个小块的大气光值,即选取暗通道先验中99.9%的像素点中最亮的像素值。
3. 估计每个小块的透射率,即根据暗通道先验和大气光值计算。
4. 修复每个小块的颜色失真,即将每个像素点的颜色值除以该像素点的透射率,再加上大气光值。
5. 将所有小块的图像块合并,取平均值,并输出结果。
需要注意的是,该代码对图像大小有限制,如果输入的图像大小不是15的倍数,则可能会出现错误。此外,该代码对输入图像的亮度和对比度要求较高,如果输入的图像过暗或过亮,则可能会出现不良效果。